Thuật ngữ AI
Từ điển đầy đủ về Trí tuệ nhân tạo
Streaming Semi-Supervised Learning
Machine learning approach that combines labeled and unlabeled data in real-time to improve models when labels are scarce or expensive to obtain.
Progressive Self-Labeling
Technique where the model generates its own labels for unlabeled data incrementally, based on an adaptive confidence threshold.
Streaming Co-Training
Method where multiple classifiers mutually train each other in streaming, each teaching others from the data it classifies with the most confidence.
Streaming Label Propagation Classification
Algorithm that propagates labels through a similarity graph constructed dynamically as new data arrives in the stream.
Online Contrastive Learning
Semi-supervised approach that learns robust representations by maximizing consistency between different augmentations of the same data in real-time.
Adaptive Pseudo-Labeling
Strategy for assigning labels to unlabeled data with dynamic confidence thresholds that adjust according to the streaming data distribution.
Dynamic Tri-Training
Extension of tri-training where three classifiers mutually train each other in real-time, with continuous update mechanisms to handle data streams.
Incremental Graph Learning
Progressive construction and exploitation of a relationship graph between instances to propagate information from labeled to unlabeled data in streaming.
Semi-supervised temporal cross-validation
Evaluation technique that respects the chronological order of data while leveraging unlabeled information to validate streaming models.
Concept drift in semi-supervised learning
Detection and adaptation to changes in data distribution or conceptual relationships in a continuous streaming semi-supervised context.
Uncertainty sampling in streaming
Active selection of the most informative instances to label among continuously arriving unlabeled data to maximize learning efficiency.
Distributed consensus learning
Approach where multiple streaming agents or models reach consensus on labels for unannotated data through weighted voting mechanisms.
Semi-supervised funnel method
Progressive filtering strategy for unlabeled data where only a fraction of the most reliable data is used for training at each stage of the stream.
Semi-supervised reinforcement learning in streaming
Combination of reinforcement learning with semi-supervised techniques to improve learning policy by leveraging unlabeled data in real-time.
Incremental multi-view learning
Extension of co-training where different representations or views of the same data are progressively used to enhance semi-supervised streaming learning.
Temporal consistency of pseudo-labels
Principle ensuring stability of predicted labels for similar instances arriving at close time intervals in the data stream.
Adaptive Semi-Supervised Ensemble Learning
Dynamic combination of multiple semi-supervised models that continuously adapt to new data to improve prediction robustness.